CN116882031B - Building model construction method and system based on point cloud - Google Patents
Building model construction method and system based on point cloud Download PDFInfo
- Publication number
- CN116882031B CN116882031B CN202311123548.0A CN202311123548A CN116882031B CN 116882031 B CN116882031 B CN 116882031B CN 202311123548 A CN202311123548 A CN 202311123548A CN 116882031 B CN116882031 B CN 116882031B
- Authority
- CN
- China
- Prior art keywords
- points
- point
- small
- value
- point cloud
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000010276 construction Methods 0.000 title claims abstract description 13
- 238000000034 method Methods 0.000 claims abstract description 36
- 238000013507 mapping Methods 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims description 18
- 230000011218 segmentation Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 7
- 108091028732 Concatemer Proteins 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000011161 development Methods 0.000 description 4
- 230000018109 developmental process Effects 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 230000032683 aging Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000005034 decoration Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012876 topography Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Geometry (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Structural Engineering (AREA)
- Computational Mathematics (AREA)
- Civil Engineering (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Architecture (AREA)
- Computer Graphics (AREA)
- Software Systems (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention provides a building model construction method and system based on point cloud, wherein a first small cuboid and a second small cuboid are obtained according to building point cloud data; respectively taking the points in each first small cuboid and the points in each second small cuboid as input of a PointNet model to obtain a characteristic value of each first small cuboid and a characteristic value of each second small cuboid, determining a first characteristic and a second characteristic according to the first small cuboid and the second small cuboid corresponding to the points in the point cloud, fusing the first characteristic, the second characteristic, the global characteristic of the point cloud and the characteristics of the points to obtain comprehensive characteristics of the points, and dividing the point cloud according to the comprehensive characteristics of each point in the point cloud; and carrying out surface reconstruction on the segmented point cloud, and obtaining a three-dimensional model of the building after texture mapping. The method can effectively improve the accuracy of building three-dimensional model reconstruction.
Description
Technical Field
The invention relates to the field of building models, in particular to a building model construction method and system based on point cloud.
Background
The building model has important significance in the fields of ancient building protection, digital cities, smart cities and the like, and provides powerful support for sustainable development and cultural inheritance of the fields. Ancient architecture often faces aging, corrosion, natural disasters and other challenges, and building models can help to make scientific protection plans, so that inheritance of historical culture is realized. The application of the building model is not only to record the appearance of a historical building, but also to present the internal structure, material characteristics and spatial layout of the building through a digitizing technology. The digital models can provide reliable basis for repair and maintenance work, so that the repair process is more accurate and effective. In addition, through the digital modeling technology, the city planner can simulate the conditions of city development, traffic flow, population distribution and the like, so that city planning and resource allocation are better carried out. The building model can assist decision makers in knowing the spatial distribution of cities, improve the intelligent level of city management, optimize the running efficiency of cities and realize sustainable city development. Through real-time monitoring and data analysis, the service conditions, energy consumption and the like of the building can be accurately managed, the optimal allocation of resources is realized, and the sustainable development of cities is promoted.
LiDAR (Light Detection and Ranging) can measure the topography, shape, height and position of a building, etc. with high accuracy, providing accurate basic data for building models, which are mainly generated by emitting laser pulses and measuring return time, a large amount of three-dimensional point cloud data including outline, shape, roof, etc. of the building. For fine modeling, capturing details of a building, such as decoration of facades, windows, doors, etc. However, the existing method for generating the three-dimensional model of the building based on the point cloud is complex in calculation and low in precision.
Disclosure of Invention
In order to solve the problems, the invention provides a building model construction method based on point cloud, which comprises the following steps:
acquiring LiDAR point cloud data of a building, preprocessing the point cloud data, acquiring a first small cuboid and a second small cuboid according to the point cloud data, processing points in the first small cuboid to enable the points of the point cloud contained in the first small cuboid to be a first fixed value, and similarly processing the points in the second small cuboid to enable the points of the point cloud contained in the second small cuboid to be a second fixed value;
respectively taking the points in each first small cuboid and the points in each second small cuboid as input of a PointNet model to obtain a characteristic value of each first small cuboid and a characteristic value of each second small cuboid, determining a first characteristic and a second characteristic according to the first small cuboid and the second small cuboid corresponding to the points in the point cloud, fusing the first characteristic, the second characteristic, the global characteristic of the point cloud and the characteristics of the points to obtain comprehensive characteristics of the points, and dividing the point cloud according to the comprehensive characteristics of each point in the point cloud;
And carrying out surface reconstruction on the segmented point cloud, and obtaining a three-dimensional model of the building after texture mapping.
Preferably, the first small cuboid and the second small cuboid are obtained according to the point cloud data, specifically:
obtaining a maximum x value, a maximum y value and a maximum z value in all point cloud data, and obtaining a minimum x value, a minimum y value and a minimum z value in all point cloud data, wherein the minimum z value, the maximum y value, the maximum z value, the minimum x value, the minimum y value and the minimum z value are a cuboid according to the maximum x value, the maximum y value, the maximum z value and the minimum x value;
dividing the cuboids to obtain a plurality of first small cuboids with the same size, filtering out the first small cuboids without point clouds, calculating the point cloud point closest to the center of the first small cuboids in the first small cuboids, taking the closest point cloud point as a key point, and determining the length, width and height of the second small cuboids according to the distribution condition of the point clouds in the first small cuboids where the key points are located.
Preferably, the determining the length, width and height of the second small cuboid according to the distribution condition of the point cloud of the first small cuboid where the key point is located specifically is:
calculating the mean square error of the point cloud x value of the first small cuboidMean square error of y value->Mean square error of z value->；
According to the formula Calculating to obtain the length, width and height of the second small cuboid,、/>for adjusting parameters, and->I is x, y or z, < >>Indicating the length, width or height of the first small cuboid,the second small cuboid is long, wide or high.
Preferably, the processing the points in the first small cuboid makes the point number of the point cloud contained in the first small cuboid be a first fixed value, specifically:
judging the relation between the number of the midpoints of the first small cuboid and the first fixed value;
if the number is larger than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and extracting the points with the first fixed value from the sorted points at equal intervals to serve as points of point clouds contained in the first small cuboid;
if the number is smaller than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and inserting a plurality of points into the sorted points so that the number of point clouds after the points are inserted is the first fixed value.
Preferably, the inserting a plurality of points in the ordered points makes the number of point clouds after the inserting of the points be the first fixed value, specifically:
S11, calculating the difference value between the first fixed value and the number, and if the difference value is smaller than the number, executing S12; if the difference is equal to the number, copying each point in the first small cuboid once; if the difference is greater than the number, copying each point in the first small cuboid n times, and then executing S12;
s12, selecting m points closest to the key point from the ordered points, and copying the closest m points so that the number of point clouds after the points are inserted is the first fixed value;
wherein,，/>and (3) representing the difference value, wherein N represents the number, and m is a positive integer.
Preferably, the first feature and the second feature are determined according to a first small cuboid and a second small cuboid corresponding to points in the point cloud, specifically:
acquiring a first small cuboid and a second small cuboid corresponding to points in the point cloud, if one point is in a plurality of second small cuboids, calculating the distance between the point and the center of each second small cuboid, and taking the second small cuboid with the shortest distance as the second small cuboid corresponding to the point;
taking points in the first small cuboid as input of a PointNet model, and taking global features output by the PointNet model as first features;
The points in the second small cuboid are taken as input of the PointNet model, and the global feature output by the PointNet model is taken as a second feature.
Preferably, the merging the first feature, the second feature, the global feature of the point cloud and the feature of the point to obtain the comprehensive feature of the point specifically includes:
acquiring key points of each first small cuboid, taking all the key points as input of a PointNet model, and taking global features output by the PointNet model as global features of point cloud; the key points are point cloud points in the first small cuboid, which are closest to the center of the first small cuboid;
the method comprises the steps of encoding point information to obtain point characteristics, and performing concatemer operation on the point characteristics, the first characteristics, the second characteristics and the global characteristics of point cloud to obtain point comprehensive characteristics.
In another aspect, the present invention provides a building model construction system based on a three-dimensional point cloud, the system comprising:
the point cloud segmentation module is used for acquiring LiDAR point cloud data of a building, preprocessing the point cloud data, acquiring a first small cuboid and a second small cuboid according to the point cloud data, processing points in the first small cuboid to enable the points of the point cloud contained in the first small cuboid to be a first fixed value, and similarly processing the points in the second small cuboid to enable the points of the point cloud contained in the second small cuboid to be a second fixed value;
The point cloud segmentation module is used for respectively taking the points in each first small cuboid and the points in each second small cuboid as the input of the PointNet model to obtain the characteristic value of each first small cuboid and the characteristic value of each second small cuboid, determining the first characteristic and the second characteristic according to the first small cuboid and the second small cuboid corresponding to the points in the point cloud, and fusing the first characteristic, the second characteristic, the global characteristic of the point cloud and the characteristic of the point to obtain the comprehensive characteristic of the point, and segmenting the point cloud according to the comprehensive characteristic of each point in the point cloud;
the model construction module is used for carrying out surface reconstruction on the segmented point cloud and obtaining a three-dimensional model of the building after texture mapping.
Preferably, the first small cuboid and the second small cuboid are obtained according to the point cloud data, specifically:
obtaining a maximum x value, a maximum y value and a maximum z value in all point cloud data, and obtaining a minimum x value, a minimum y value and a minimum z value in all point cloud data, wherein the minimum z value, the maximum y value, the maximum z value, the minimum x value, the minimum y value and the minimum z value are a cuboid according to the maximum x value, the maximum y value, the maximum z value and the minimum x value;
dividing the cuboids to obtain a plurality of first small cuboids with the same size, filtering out the first small cuboids without point clouds, calculating the point cloud point closest to the center of the first small cuboids in the first small cuboids, taking the closest point cloud point as a key point, and determining the length, width and height of the second small cuboids according to the distribution condition of the point clouds in the first small cuboids where the key points are located.
Preferably, the determining the length, width and height of the second small cuboid according to the distribution condition of the point cloud of the first small cuboid where the key point is located specifically is:
calculating the mean square error of the point cloud x value of the first small cuboidMean square error of y value->Mean square error of z value->；
According to the formulaCalculating to obtain the length, width and height of the second small cuboid,、/>for adjusting parameters, and->I is x, y or z, < >>Indicating the length, width or height of the first small cuboid,the second small cuboid is long, wide or high.
Preferably, the processing the points in the first small cuboid makes the point number of the point cloud contained in the first small cuboid be a first fixed value, specifically:
judging the relation between the number of the midpoints of the first small cuboid and the first fixed value;
if the number is larger than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and extracting the points with the first fixed value from the sorted points at equal intervals to serve as points of point clouds contained in the first small cuboid;
if the number is smaller than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and inserting a plurality of points into the sorted points so that the number of point clouds after the points are inserted is the first fixed value.
Preferably, the inserting a plurality of points in the ordered points makes the number of point clouds after the inserting of the points be the first fixed value, specifically:
s11, calculating the difference value between the first fixed value and the number, and if the difference value is smaller than the number, executing S12; if the difference is equal to the number, copying each point in the first small cuboid once; if the difference is greater than the number, copying each point in the first small cuboid n times, and then executing S12;
s12, selecting m points closest to the key point from the ordered points, and copying the closest m points so that the number of point clouds after the points are inserted is the first fixed value;
wherein,，/>and (3) representing the difference value, wherein N represents the number, and m is a positive integer.
Preferably, the first feature and the second feature are determined according to a first small cuboid and a second small cuboid corresponding to points in the point cloud, specifically:
acquiring a first small cuboid and a second small cuboid corresponding to points in the point cloud, if one point is in a plurality of second small cuboids, calculating the distance between the point and the center of each second small cuboid, and taking the second small cuboid with the shortest distance as the second small cuboid corresponding to the point;
Taking points in the first small cuboid as input of a PointNet model, and taking global features output by the PointNet model as first features;
the points in the second small cuboid are taken as input of the PointNet model, and the global feature output by the PointNet model is taken as a second feature.
Preferably, the merging the first feature, the second feature, the global feature of the point cloud and the feature of the point to obtain the comprehensive feature of the point specifically includes:
acquiring key points of each first small cuboid, taking all the key points as input of a PointNet model, and taking global features output by the PointNet model as global features of point cloud; the key points are point cloud points in the first small cuboid, which are closest to the center of the first small cuboid;
the method comprises the steps of encoding point information to obtain point characteristics, and performing concatemer operation on the point characteristics, the first characteristics, the second characteristics and the global characteristics of point cloud to obtain point comprehensive characteristics.
Furthermore, the present invention provides a computer readable storage medium having stored thereon a computer program which, when executed, implements a method as described above.
Finally, the invention also provides an electronic device, which comprises: a memory, a processor; wherein the memory has stored thereon executable program code which, when executed by the processor, causes the processor to perform the method as described above.
Aiming at the problem of poor reconstruction effect of the existing three-dimensional building model, the method provided by the invention is used for segmenting the acquired three-dimensional building point cloud, so that the self characteristics of the points in the point cloud are considered in the segmentation, the local characteristics and the global characteristics of the point cloud are also considered, the accuracy of the point cloud segmentation can be improved, and the finally reconstructed building model is more accurate.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a first embodiment;
FIG. 2 is a partial schematic diagram of a PointNet model;
FIG. 3 is a schematic view of a rectangular parallelepiped;
fig. 4 is a schematic diagram of the relationship between the first small cuboid and the second small cuboid.
Detailed Description
In this document, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In a first embodiment, the present invention provides a building model construction method based on point cloud, as shown in fig. 1, the method includes the following steps:
s1, liDAR point cloud data of a building is obtained, the point cloud data is preprocessed, a first small cuboid and a second small cuboid are obtained according to the point cloud data, points in the first small cuboid are processed so that the points of the point cloud contained in the first small cuboid are a first fixed value, and the points in the second small cuboid are processed in the same way so that the points of the point cloud contained in the second small cuboid are a second fixed value;
LiDAR (Light Detection and Ranging) techniques use laser pulses to measure the distance of an object surface, thereby generating high density point cloud data. Point cloud data of a building is acquired by LiDAR, which is a collection of points representing the surface of an object in a three-dimensional space, consisting of a large number of discrete points, each having three-dimensional coordinates and possibly other attributes, such as reflection intensity, normal vector, etc., but the points in the point cloud all include three-dimensional coordinates (x, y, z).
After the point cloud is obtained, the space in which the point cloud is located is divided into a plurality of first small cuboids and a plurality of second small cuboids, and in a preferred embodiment, the length, width and height of the second small cuboids are respectively larger than the length, width and height of the first small cuboids, and the first small cuboids are all or partially located in the second small cuboids.
Because the positions of the first small cuboids are different, the number of the point clouds contained in the first small cuboids is different, in order to facilitate subsequent processing, the points in the first small cuboids are processed so that the number of the point clouds contained in the first small cuboids is a first fixed value, and similarly, the points in the second small cuboids are processed so that the number of the point clouds contained in the second small cuboids is a second fixed value, wherein the second fixed value is larger than the first fixed value, and the first fixed value and the second fixed value are positive integers.
S2, respectively taking the points in each first small cuboid and the points in each second small cuboid as the input of a PointNet model to obtain the characteristic value of each first small cuboid and the characteristic value of each second small cuboid, determining the first characteristic and the second characteristic according to the first small cuboid and the second small cuboid corresponding to the points in the point cloud, and fusing the first characteristic, the second characteristic, the global characteristic of the point cloud and the characteristics of the points to obtain the comprehensive characteristics of the points, and dividing the comprehensive characteristics of each point in the point cloud according to the comprehensive characteristics of the points;
PointNet is a machine learning model used for processing tasks related to point cloud data, such as classifying point clouds and segmenting point clouds. In the PointNet model, it is divided into two parts, one is a classification network and the other is a segmentation network. However, both networks involve global features (global features) that are obtained after maximum pooling, i.e., features that are output after Max Pool are global features, as shown in fig. 2.
After the step S1, the point numbers of the point clouds contained in each first small cube are the same, the point numbers of the point clouds contained in each second small cube are also all thought of, for each first small cube, all the points in the point clouds are used as input of a PointNet model, and the characteristic values corresponding to the first small cubes are obtained after the maximum pooling layer, and similarly, the characteristic values corresponding to the second small cubes can be obtained.
A point in the point cloud corresponds to a first small cube and one or more second small cubes, and a first feature and a second feature are determined according to the corresponding first small cubes and the corresponding second small cubes, wherein the first feature and the second feature reflect local position information of the point in the point cloud; the global features of the point cloud reflect the global features of the whole point cloud, the self features of the point reflect the self features of the point, the first features and the second features are fused with the global features of the point cloud and the self features of the point cloud to obtain comprehensive features of the point, and then the point cloud is segmented by using a PointNet segmentation network part. In another embodiment, after downsampling the composite features, MLP is used to semantically segment the point cloud, dividing the point cloud of the building into walls, windows, roofs, etc.
The self-characteristics of the points at least comprise (x, y, z) coordinate information of the points, and can also comprise other information, such as normal vectors, reflection intensity and the like.
And S3, carrying out surface reconstruction on the segmented point cloud, and obtaining a three-dimensional model of the building after texture mapping.
The three-dimensional model has various representation modes, such as point cloud, mesh (grid), voxels and the like, wherein the voxels are mainly applied to medical images, and compared with the point cloud, the grid is more suitable for rendering, visualization and the like, and after the point cloud is segmented, the surface reconstruction mode of the point cloud comprises but is not limited to grid reconstruction, spline curve/surface fitting, poisson reconstruction and the like. After the surface is reconstructed, texture mapping is carried out, namely, a three-dimensional model of the building is obtained through texture mapping, and then the building model with reality can be obtained through texture mapping, so that the method has great significance for ancient building protection, ancient building restoration and smart city.
In addition to global information, the points in the point cloud are related to local information of each point, and in a specific embodiment, the first small cuboid and the second small cuboid are obtained according to the point cloud data, specifically:
Obtaining a maximum x value, a maximum y value and a maximum z value in all point cloud data, and obtaining a minimum x value, a minimum y value and a minimum z value in all point cloud data, wherein the minimum z value, the maximum y value, the maximum z value, the minimum x value, the minimum y value and the minimum z value are a cuboid according to the maximum x value, the maximum y value, the maximum z value and the minimum x value;
in a more specific embodiment, before the maximum x value, the maximum y value and the maximum z value in all the point cloud data are obtained, the point cloud image of the building is selected to be in the front view direction, so that a cuboid which can tightly surround all the point clouds is obtained. In another embodiment, if the building includes a plurality of buildings, the point cloud may be divided first to obtain point clouds of a plurality of individual buildings, and the cutting of each building adopts a bounding box mode.
The points in the point cloud have position coordinate information (x, y, z), and a cube can be obtained through the maximum value, the minimum value, the maximum value and the minimum value of the x value, specifically, the maximum value of the x value is obtained, then a plane parallel to the yz plane is obtained at the maximum value, and similarly, other five planes can be obtained, and a space formed by the five planes is the cuboid, as shown in fig. 3.
Dividing the cuboids to obtain a plurality of first small cuboids with the same size, filtering out the first small cuboids without point clouds, calculating the point cloud point closest to the center of the first small cuboids in the first small cuboids, taking the closest point cloud point as a key point, and determining the length, width and height of the second small cuboids according to the distribution condition of the point clouds in the first small cuboids where the key points are located.
All points in the point cloud are in the cuboid, the cuboid is divided to obtain a plurality of first small cuboids with equal sizes, the number of the first small cuboids is obtained in a plurality of modes, and preferably, the cuboid is cut in a continuous iteration mode until the number of the point cloud which is most included in the obtained first small cuboid is smaller than a preset value. The continuous iteration means that the cuboid is cut into four first small cuboids for the first time, if the four first small cuboids do not meet the requirement, each of the four first small cuboids is subdivided into four, and the above process is continuously repeated until the obtained point cloud data contained in all the first small cuboids is smaller than a preset value.
And filtering the first small cuboids without the point cloud midpoint, namely deleting the first small cuboids, acquiring the center point of each first small cuboid for the first small cuboids with the point cloud, finding out the point cloud point which is positioned in the first small cuboids and is closest to the center point, taking the point cloud point as a key point, and determining the length, width and height of the second small cuboids according to the distribution condition of the point cloud of the first small cuboids with the key points.
The number of the first small rectangular parallelepiped is the same as the number of the second small rectangular parallelepiped, and of course, the first small rectangular parallelepiped refers to the filtered first small rectangular parallelepiped. And there may be an overlapping region in the space between the second small cuboids.
In a specific embodiment, the determining, according to the cloud distribution condition of the points of the first small cuboid where the key points are located, the length, width and height of the second small cuboid is specifically:
calculating the mean square error of the point cloud x value of the first small cuboidMean square error of y value->Mean square error of z value->；
According to the formulaCalculating to obtain the length, width and height of the second small cuboid,、/>for adjusting parameters, and->I is x, y or z, < >>Indicating the length, width or height of the first small cuboid,the second small cuboid is long, wide or high.
The mean square error is the square root of the variance, the deviation degree between the data can be well described, and the calculation formula of the mean square error is as follows:wherein->Indicating the desire. The distribution of the points in the first microcubes can be obtained by means of the mean square error, if the points in the first microcubes are more concentrated in one plane, for example in one limit, all points being in the plane of the first microcubes with x=1, then- >And->All zero, in which case the second cube may be a very flat cube.
Since the second small cube mainly acquires the distribution characteristics of point clouds in a larger area, the method is based on the formulaAnd obtaining the length, width and height of the second small cube. For example, when i is x, the above formula becomes a formula for calculating the length, i.e., +.>Likewise, the formulas for calculating the width and height of the second small cube can be obtained as: />、/>. The center of the second small cube is the key point, as shown in fig. 4.
The number of points of the point cloud contained in each first small cube and each second small cube is different, some points contain 50 points, and some points may contain 5 points, in a specific embodiment, the points in the first small cube are processed so that the points of the point cloud contained in the first small cube are a first fixed value, and specifically:
judging the relation between the number of the midpoints of the first small cuboid and the first fixed value;
if the number is larger than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and extracting the points with the first fixed value from the sorted points at equal intervals to serve as points of point clouds contained in the first small cuboid;
At this time, the number of points contained in the first small cube is large, and extraction is required, specifically, all the points in the first small cube are ordered according to the distance from the key point, and then the first fixed number of points are extracted from the ordered points at equal intervals. For example, ten points after sorting are respectively 1, 2, 3, 4, 5, 6, 7, 8, 9 and 10, and 1, 4, 7 and 10 are extracted assuming that the first fixed value is 4. If the equal interval cannot be strictly set, the points furthest apart are discarded so that the equal interval is satisfied.
If the number is smaller than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and inserting a plurality of points into the sorted points so that the number of point clouds after the points are inserted is the first fixed value.
If the number of points contained in the first small cube is small, interpolation is needed, and a plurality of points are inserted into the ordered points, so that the number of points contained in the first small cube meets a first fixed value. In a specific embodiment, the inserting a plurality of points in the ordered points makes the number of point clouds after the point is inserted be the first fixed value, specifically:
S11, calculating the difference value between the first fixed value and the number, and if the difference value is smaller than the number, executing S12; if the difference is equal to the number, copying each point in the first small cuboid once; if the difference is greater than the number, copying each point in the first small cuboid n times, and then executing S12;
s12, selecting m points closest to the key point from the ordered points, and copying the closest m points so that the number of point clouds after the points are inserted is the first fixed value;
wherein,，/>and (3) representing the difference value, wherein N represents the number, and m is a positive integer.
In interpolation, if the difference between the number of points contained in the first cube and the first fixed value is small (that is, the difference between the number of points is smaller than or equal to the number of points contained in the first cube), m points closest to the key point are selected for copying, and the sum of the number of points originally contained in the first cube and the m points obtained by copying is the first fixed value.
If the difference between the first fixed value and the points contained in the first cube is large, the points needing to be copied are large, at the moment, the points originally contained in the first cube are copied for n times, then the difference is reduced, and then a plurality of points closest to the key point are carried out, so that the point number of the final result is equal to the first fixed value.
Further, if the number of points contained in the first small cube is the same as the first fixed value, no processing is necessary.
When the point cloud is segmented, not only the characteristics of the point cloud but also the characteristics of peripheral points need to be referred to, and the first characteristics and the second characteristics are determined according to the first small cuboid and the second small cuboid corresponding to the points in the point cloud, specifically:
acquiring a first small cuboid and a second small cuboid corresponding to points in the point cloud, if one point is in a plurality of second small cuboids, calculating the distance between the point and the center of each second small cuboid, and taking the second small cuboid with the shortest distance as the second small cuboid corresponding to the point;
because when cutting the cuboid, first little cuboid is next, does not have the overlapping area, does not have the clearance moreover, every point all corresponds a first cuboid like this, but the center of second cuboid and the center of first cuboid are different, and the size is also different, and the condition that overlaps can appear in the second cuboid.
When a first small cuboid corresponding to a point in the point cloud is determined, the first small cuboid can be uniquely determined, but some points fall into a plurality of second small cuboids, and based on the first small cuboid, the second small cuboid closest to the center of the second small cuboid is selected as the second small cuboid corresponding to the point.
Taking points in the first small cuboid as input of a PointNet model, and taking global features output by the PointNet model as first features;
the points in the second small cuboid are taken as input of the PointNet model, and the global feature output by the PointNet model is taken as a second feature.
Then taking the point in the first small cuboid as the input of the PointNet model, and taking the global feature output after Max Pool as the first feature; similarly, the point in the second small rectangular parallelepiped is taken as the input of the PointNet model, and the global feature output after Max Pool is taken as the second feature.
In a specific embodiment, the merging of the first feature, the second feature, the global feature of the point cloud and the feature of the point to obtain the comprehensive feature of the point specifically includes:
acquiring key points of each first small cuboid, taking all the key points as input of a PointNet model, and taking global features output by the PointNet model as global features of point cloud; the key points are point cloud points in the first small cuboid, which are closest to the center of the first small cuboid;
the number of points in the point cloud of a building is often millions, if the points are used for calculating global features, the calculated amount is large, and the key points of each first small cuboid are used as input of a PointNet model, and the global features output after Max Pool are used as global features of the point cloud.
The method comprises the steps of encoding point information to obtain point characteristics, and performing concatemer operation on the point characteristics, the first characteristics, the second characteristics and the global characteristics of point cloud to obtain point comprehensive characteristics.
There are various ways to encode the point information, one way is to use the encoding way of the PointNet model, namely, each point is used as input to obtain the vector of the point input Max Pool, and the vector is used as the encoding result of the point information to obtain the point characteristics; another way is to use a separate encoding to embed the information of the points into a fixed length feature, e.g. 2048.
And finally, carrying out concat operation on the characteristics of the points, the first characteristics, the second characteristics and the global characteristics of the point cloud in sequence to obtain the comprehensive characteristics of the points. And classifying the comprehensive characteristics of the points through MLP or FFN to obtain the classification of the points, thereby realizing the semantic segmentation of the point cloud.
In a second embodiment, the present invention provides a building model construction system based on a point cloud, the system including the following modules:
the point cloud segmentation module is used for acquiring LiDAR point cloud data of a building, preprocessing the point cloud data, acquiring a first small cuboid and a second small cuboid according to the point cloud data, processing points in the first small cuboid to enable the points of the point cloud contained in the first small cuboid to be a first fixed value, and similarly processing the points in the second small cuboid to enable the points of the point cloud contained in the second small cuboid to be a second fixed value;
The point cloud segmentation module is used for respectively taking the points in each first small cuboid and the points in each second small cuboid as the input of the PointNet model to obtain the characteristic value of each first small cuboid and the characteristic value of each second small cuboid, determining the first characteristic and the second characteristic according to the first small cuboid and the second small cuboid corresponding to the points in the point cloud, and fusing the first characteristic, the second characteristic, the global characteristic of the point cloud and the characteristic of the point to obtain the comprehensive characteristic of the point, and segmenting the point cloud according to the comprehensive characteristic of each point in the point cloud;
the model construction module is used for carrying out surface reconstruction on the segmented point cloud and obtaining a three-dimensional model of the building after texture mapping.
Preferably, the first small cuboid and the second small cuboid are obtained according to the point cloud data, specifically:
obtaining a maximum x value, a maximum y value and a maximum z value in all point cloud data, and obtaining a minimum x value, a minimum y value and a minimum z value in all point cloud data, wherein the minimum z value, the maximum y value, the maximum z value, the minimum x value, the minimum y value and the minimum z value are a cuboid according to the maximum x value, the maximum y value, the maximum z value and the minimum x value;
dividing the cuboids to obtain a plurality of first small cuboids with the same size, filtering out the first small cuboids without point clouds, calculating the point cloud point closest to the center of the first small cuboids in the first small cuboids, taking the closest point cloud point as a key point, and determining the length, width and height of the second small cuboids according to the distribution condition of the point clouds in the first small cuboids where the key points are located.
Preferably, the determining the length, width and height of the second small cuboid according to the distribution condition of the point cloud of the first small cuboid where the key point is located specifically is:
calculating the mean square error of the point cloud x value of the first small cuboidMean square error of y value->Mean square error of z value->；
According to the formulaCalculating to obtain the length, width and height of the second small cuboid,、/>for adjusting parameters, and->I is x, y or z, < >>Indicating the length, width or height of the first small cuboid,is the length of a second small cuboid,Wide or high.
Preferably, the processing the points in the first small cuboid makes the point number of the point cloud contained in the first small cuboid be a first fixed value, specifically:
judging the relation between the number of the midpoints of the first small cuboid and the first fixed value;
if the number is larger than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and extracting the points with the first fixed value from the sorted points at equal intervals to serve as points of point clouds contained in the first small cuboid;
if the number is smaller than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and inserting a plurality of points into the sorted points so that the number of point clouds after the points are inserted is the first fixed value.
Preferably, the inserting a plurality of points in the ordered points makes the number of point clouds after the inserting of the points be the first fixed value, specifically:
s11, calculating the difference value between the first fixed value and the number, and if the difference value is smaller than the number, executing S12; if the difference is equal to the number, copying each point in the first small cuboid once; if the difference is greater than the number, copying each point in the first small cuboid n times, and then executing S12;
s12, selecting m points closest to the key point from the ordered points, and copying the closest m points so that the number of point clouds after the points are inserted is the first fixed value;
wherein,，/>and (3) representing the difference value, wherein N represents the number, and m is a positive integer.
Preferably, the first feature and the second feature are determined according to a first small cuboid and a second small cuboid corresponding to points in the point cloud, specifically:
acquiring a first small cuboid and a second small cuboid corresponding to points in the point cloud, if one point is in a plurality of second small cuboids, calculating the distance between the point and the center of each second small cuboid, and taking the second small cuboid with the shortest distance as the second small cuboid corresponding to the point;
Taking points in the first small cuboid as input of a PointNet model, and taking global features output by the PointNet model as first features;
the points in the second small cuboid are taken as input of the PointNet model, and the global feature output by the PointNet model is taken as a second feature.
Preferably, the merging the first feature, the second feature, the global feature of the point cloud and the feature of the point to obtain the comprehensive feature of the point specifically includes:
acquiring key points of each first small cuboid, taking all the key points as input of a PointNet model, and taking global features output by the PointNet model as global features of point cloud; the key points are point cloud points in the first small cuboid, which are closest to the center of the first small cuboid;
the method comprises the steps of encoding point information to obtain point characteristics, and performing concatemer operation on the point characteristics, the first characteristics, the second characteristics and the global characteristics of point cloud to obtain point comprehensive characteristics.
Furthermore, the present invention provides a computer readable storage medium having stored thereon a computer program which, when executed, implements a method as described above.
Finally, the invention also provides an electronic device, which comprises: a memory, a processor; wherein the memory has stored thereon executable program code which, when executed by the processor, causes the processor to perform the method as described above.
In a third embodiment, the present invention further provides a computer readable storage medium, where a computer program is stored, where the computer program when executed by a computer implements the method according to the first embodiment.
The fourth embodiment of the present invention further provides an electronic device, including: a memory, a processor; wherein the memory has stored thereon executable program code which, when executed by the processor, causes the processor to perform the method as described in embodiment one.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by adding necessary general purpose hardware platforms, or may be implemented by a combination of hardware and software. Based on such understanding, the foregoing aspects, in essence and portions contributing to the art, may be embodied in the form of a computer program product, which may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (6)
1. The building model construction method based on the point cloud is characterized by comprising the following steps of:
acquiring LiDAR point cloud data of a building, preprocessing the point cloud data, acquiring a first small cuboid and a second small cuboid according to the point cloud data, processing points in the first small cuboid to enable the points of the point cloud contained in the first small cuboid to be a first fixed value, and similarly processing the points in the second small cuboid to enable the points of the point cloud contained in the second small cuboid to be a second fixed value;
respectively taking the points in each first small cuboid and the points in each second small cuboid as input of a PointNet model to obtain a characteristic value of each first small cuboid and a characteristic value of each second small cuboid, determining a first characteristic and a second characteristic according to the first small cuboid and the second small cuboid corresponding to the points in the point cloud, fusing the first characteristic, the second characteristic, the global characteristic of the point cloud and the characteristics of the points to obtain comprehensive characteristics of the points, and dividing the point cloud according to the comprehensive characteristics of each point in the point cloud;
Reconstructing the surface of the segmented point cloud, and obtaining a three-dimensional model of the building after texture mapping;
the processing of the points in the first small cuboid enables the points of the point cloud contained in the first small cuboid to be a first fixed value, specifically:
judging the relation between the number of the midpoints of the first small cuboid and the first fixed value;
the method comprises the steps of obtaining a first small cuboid and a second small cuboid according to point cloud data, wherein the first small cuboid and the second small cuboid are specifically:
obtaining a maximum x value, a maximum y value and a maximum z value in all point cloud data, and obtaining a minimum x value, a minimum y value and a minimum z value in all point cloud data, wherein the minimum z value, the maximum y value, the maximum z value, the minimum x value, the minimum y value and the minimum z value are a cuboid according to the maximum x value, the maximum y value, the maximum z value and the minimum x value;
dividing the cuboids to obtain a plurality of first small cuboids with the same size, filtering out the first small cuboids without point clouds, calculating the point cloud point closest to the center of the first small cuboids in the first small cuboids, taking the closest point cloud point as a key point, and determining the length, width and height of the second small cuboids according to the distribution condition of the point clouds in the first small cuboids where the key points are located;
if the number is larger than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and extracting the points with the first fixed value from the sorted points at equal intervals to serve as points of point clouds contained in the first small cuboid;
If the number is smaller than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and inserting a plurality of points into the sorted points so that the number of point clouds after the points are inserted is the first fixed value;
the method comprises the steps of determining a first characteristic and a second characteristic according to a first small cuboid and a second small cuboid corresponding to points in a point cloud, wherein the first characteristic and the second characteristic are specifically as follows:
acquiring a first small cuboid and a second small cuboid corresponding to points in the point cloud, if one point is in a plurality of second small cuboids, calculating the distance between the point and the center of each second small cuboid, and taking the second small cuboid with the shortest distance as the second small cuboid corresponding to the point;
taking points in the first small cuboid as input of a PointNet model, and taking global features output by the PointNet model as first features;
taking points in the second small cuboid as input of the PointNet model, and taking global features output by the PointNet model as second features;
the method comprises the steps of fusing the first feature, the second feature, the global feature of the point cloud and the feature of the point to obtain the comprehensive feature of the point, and specifically comprises the following steps:
acquiring key points of each first small cuboid, taking all the key points as input of a PointNet model, and taking global features output by the PointNet model as global features of point cloud; the key points are point cloud points in the first small cuboid, which are closest to the center of the first small cuboid;
The method comprises the steps of encoding point information to obtain point characteristics, and performing concatemer operation on the point characteristics, the first characteristics, the second characteristics and the global characteristics of point cloud to obtain point comprehensive characteristics.
2. The method of claim 1, wherein the determining the length, width and height of the second small cuboid according to the distribution of the point clouds of the first small cuboid where the key points are located specifically comprises:
calculating the mean square error of the point cloud x value of the first small cuboidMean square error of y value->Mean square error of z value->；
According to the formulaCalculating the length, width and height of the second small cuboid, wherein +.>、/>For adjusting parameters, and->I is x, y or z, < >>Representing the length, width or height of the first small cuboid, +.>The second small cuboid is long, wide or high.
3. The method according to claim 1, wherein the inserting a plurality of points in the ordered points makes the number of point clouds after the inserting of the points the first fixed value, specifically:
s11, calculating the difference value between the first fixed value and the number, and if the difference value is smaller than the number, executing S12; if the difference is equal to the number, copying each point in the first small cuboid once; if the difference is greater than the number, copying each point in the first small cuboid n times, and then executing S12;
S12, selecting m points closest to the key point from the ordered points, and copying the closest m points so that the number of point clouds after the points are inserted is the first fixed value;
wherein,，/>and (3) representing the difference value, wherein N represents the number, and m is a positive integer.
4. A point cloud based building model construction system, the system comprising the following modules:
the point cloud segmentation module is used for acquiring LiDAR point cloud data of a building, preprocessing the point cloud data, acquiring a first small cuboid and a second small cuboid according to the point cloud data, processing points in the first small cuboid to enable the points of the point cloud contained in the first small cuboid to be a first fixed value, and similarly processing the points in the second small cuboid to enable the points of the point cloud contained in the second small cuboid to be a second fixed value;
the point cloud segmentation module is used for respectively taking the points in each first small cuboid and the points in each second small cuboid as the input of the PointNet model to obtain the characteristic value of each first small cuboid and the characteristic value of each second small cuboid, determining the first characteristic and the second characteristic according to the first small cuboid and the second small cuboid corresponding to the points in the point cloud, and fusing the first characteristic, the second characteristic, the global characteristic of the point cloud and the characteristic of the point to obtain the comprehensive characteristic of the point, and segmenting the point cloud according to the comprehensive characteristic of each point in the point cloud;
The model construction module is used for carrying out surface reconstruction on the segmented point cloud and obtaining a three-dimensional model of the building after texture mapping;
the processing of the points in the first small cuboid enables the points of the point cloud contained in the first small cuboid to be a first fixed value, specifically:
judging the relation between the number of the midpoints of the first small cuboid and the first fixed value;
the method comprises the steps of obtaining a first small cuboid and a second small cuboid according to point cloud data, wherein the first small cuboid and the second small cuboid are specifically:
obtaining a maximum x value, a maximum y value and a maximum z value in all point cloud data, and obtaining a minimum x value, a minimum y value and a minimum z value in all point cloud data, wherein the minimum z value, the maximum y value, the maximum z value, the minimum x value, the minimum y value and the minimum z value are a cuboid according to the maximum x value, the maximum y value, the maximum z value and the minimum x value;
dividing the cuboids to obtain a plurality of first small cuboids with the same size, filtering out the first small cuboids without point clouds, calculating the point cloud point closest to the center of the first small cuboids in the first small cuboids, taking the closest point cloud point as a key point, and determining the length, width and height of the second small cuboids according to the distribution condition of the point clouds in the first small cuboids where the key points are located;
if the number is larger than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and extracting the points with the first fixed value from the sorted points at equal intervals to serve as points of point clouds contained in the first small cuboid;
If the number is smaller than the first fixed value, calculating the distance between each point in the first small cuboid and the key point, sorting the points according to the distance, and inserting a plurality of points into the sorted points so that the number of point clouds after the points are inserted is the first fixed value;
the method comprises the steps of determining a first characteristic and a second characteristic according to a first small cuboid and a second small cuboid corresponding to points in a point cloud, wherein the first characteristic and the second characteristic are specifically as follows:
acquiring a first small cuboid and a second small cuboid corresponding to points in the point cloud, if one point is in a plurality of second small cuboids, calculating the distance between the point and the center of each second small cuboid, and taking the second small cuboid with the shortest distance as the second small cuboid corresponding to the point;
taking points in the first small cuboid as input of a PointNet model, and taking global features output by the PointNet model as first features;
taking points in the second small cuboid as input of the PointNet model, and taking global features output by the PointNet model as second features;
the method comprises the steps of fusing the first feature, the second feature, the global feature of the point cloud and the feature of the point to obtain the comprehensive feature of the point, and specifically comprises the following steps:
acquiring key points of each first small cuboid, taking all the key points as input of a PointNet model, and taking global features output by the PointNet model as global features of point cloud; the key points are point cloud points in the first small cuboid, which are closest to the center of the first small cuboid;
The method comprises the steps of encoding point information to obtain point characteristics, and performing concatemer operation on the point characteristics, the first characteristics, the second characteristics and the global characteristics of point cloud to obtain point comprehensive characteristics.
5. A computer readable storage medium, characterized in that the computer program is stored on the readable storage medium, which computer program, when executed, implements the method according to any of claims 1-3.
6. An electronic device, the electronic device comprising: a memory, a processor; wherein the memory has stored thereon executable program code which, when executed by the processor, causes the processor to perform the method of any of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311123548.0A CN116882031B (en) | 2023-09-01 | 2023-09-01 | Building model construction method and system based on point cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311123548.0A CN116882031B (en) | 2023-09-01 | 2023-09-01 | Building model construction method and system based on point cloud |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116882031A CN116882031A (en) | 2023-10-13 |
| CN116882031B true CN116882031B (en) | 2023-11-17 |
Family
ID=88260830
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311123548.0A Active CN116882031B (en) | 2023-09-01 | 2023-09-01 | Building model construction method and system based on point cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116882031B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114187310A (en) * | 2021-11-22 | 2022-03-15 | 华南农业大学 | Large-scale point cloud segmentation method based on octree and PointNet ++ network |
| CN115115839A (en) * | 2022-07-28 | 2022-09-27 | 南京工业大学 | Building indoor point cloud segmentation method based on local feature enhanced PointNet + + network |
| WO2022252274A1 (en) * | 2021-05-31 | 2022-12-08 | 北京理工大学 | Point cloud segmentation and virtual environment generation method and apparatus based on pointnet network |
| CN115761172A (en) * | 2022-10-10 | 2023-03-07 | 哈尔滨工程大学 | Single building three-dimensional reconstruction method based on point cloud semantic segmentation and structure fitting |
| WO2023035822A1 (en) * | 2021-09-13 | 2023-03-16 | 上海芯物科技有限公司 | Target detection method and apparatus, and device and storage medium |
| CN116310192A (en) * | 2022-12-28 | 2023-06-23 | 江苏省测绘研究所 | Urban building three-dimensional model monomer reconstruction method based on point cloud |
-
2023
- 2023-09-01 CN CN202311123548.0A patent/CN116882031B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022252274A1 (en) * | 2021-05-31 | 2022-12-08 | 北京理工大学 | Point cloud segmentation and virtual environment generation method and apparatus based on pointnet network |
| WO2023035822A1 (en) * | 2021-09-13 | 2023-03-16 | 上海芯物科技有限公司 | Target detection method and apparatus, and device and storage medium |
| CN114187310A (en) * | 2021-11-22 | 2022-03-15 | 华南农业大学 | Large-scale point cloud segmentation method based on octree and PointNet ++ network |
| CN115115839A (en) * | 2022-07-28 | 2022-09-27 | 南京工业大学 | Building indoor point cloud segmentation method based on local feature enhanced PointNet + + network |
| CN115761172A (en) * | 2022-10-10 | 2023-03-07 | 哈尔滨工程大学 | Single building three-dimensional reconstruction method based on point cloud semantic segmentation and structure fitting |
| CN116310192A (en) * | 2022-12-28 | 2023-06-23 | 江苏省测绘研究所 | Urban building three-dimensional model monomer reconstruction method based on point cloud |
Non-Patent Citations (2)
| Title |
|---|
| 基于DGCNN语义分割的倾斜摄影测量三维点云建筑物分类;韩姗姗;黄远程;白穆;;测绘标准化(第03期);全文 * |
| 面向点云的三维物体识别方法综述;郝雯;王映辉;宁小娟;梁玮;石争浩;;计算机科学(第09期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116882031A (en) | 2023-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2024077812A1 (en) | Single building three-dimensional reconstruction method based on point cloud semantic segmentation and structure fitting | |
| Vo et al. | Octree-based region growing for point cloud segmentation | |
| Xu et al. | Reconstruction of scaffolds from a photogrammetric point cloud of construction sites using a novel 3D local feature descriptor | |
| Poullis et al. | Automatic reconstruction of cities from remote sensor data | |
| CN113487730B (en) | Urban three-dimensional automatic modeling method based on laser radar point cloud data | |
| CN106133756A (en) | For filtering, split and identify the system without the object in constraint environment | |
| Hu et al. | Structure‐aware 3D reconstruction for cable‐stayed bridges: A learning‐based method | |
| CN110009671B (en) | Grid curved surface reconstruction system for scene understanding | |
| Bouzas et al. | Structure-aware building mesh polygonization | |
| CN113724279B (en) | System, method, equipment and storage medium for automatically dividing traffic cells into road networks | |
| Ortega et al. | Generating 3D city models from open LiDAR point clouds: Advancing towards smart city applications | |
| Hu et al. | Geometric feature enhanced line segment extraction from large-scale point clouds with hierarchical topological optimization | |
| CN114419278A (en) | Indoor three-dimensional color grid model generation method and system | |
| Liu et al. | Generation of concise 3D building model from dense meshes by extracting and completing planar primitives | |
| Zhao et al. | A 3D modeling method for buildings based on LiDAR point cloud and DLG | |
| Zhou | 3D urban modeling from city-scale aerial LiDAR data | |
| CN116882031B (en) | Building model construction method and system based on point cloud | |
| Wang et al. | Branching the limits: Robust 3D tree reconstruction from incomplete laser point clouds | |
| CN116051980A (en) | Building identification method, system, electronic equipment and medium based on oblique photography | |
| CN114445574A (en) | Method, device and equipment for converting GeoJSON data format into three-dimensional GLB format | |
| Gonzálvez et al. | From point cloud to surface: Modeling structures in laser scanner point clouds | |
| Mansour et al. | Hierarchical SVM for Semantic Segmentation of 3D Point Clouds for Infrastructure Scenes | |
| Xiong | Reconstructing and correcting 3d building models using roof topology graphs | |
| Zhang et al. | Multi-Data UAV Images for Large Scale Reconstruction of Buildings | |
| Liu et al. | Point2Building: Reconstructing Buildings from Airborne LiDAR Point Clouds |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |